A new energy station three-dimensional GIS simulation training system based on a real scene

Abstract

The application provides a new energy station three-dimensional GIS simulation training system based on a real scene, and relates to the technical field of simulation training, and comprises: a new energy station three-dimensional scene unit, which is used for displaying operation state information, fault positioning and display, and alarm information display of each device of the new energy station, collecting fault display results, and forming a fault case library; a full-dimension quantitative evaluation unit, which is used for performing dynamic evolution type fault simulation testing on new energy station operation and maintenance personnel based on the three-dimensional real scene model and the fault case library, and performing full-dimension quantitative evaluation on the simulation testing results in combination with dynamic evolution game and interval two-type trapezoidal fuzzy sets; and a skill portrait and training strategy construction unit, which is used for constructing operation and maintenance personnel skill portraits and dynamic cooperative training strategies based on the quantitative evaluation results of the simulation testing. The application provides scientific and systematic technical support for skill optimization and emergency capability construction of a new energy station operation and maintenance team.

Description

Technical Field

[0001] This invention relates to the field of simulation training technology, and more specifically, to a three-dimensional GIS simulation training system for new energy power stations built based on real-world scenarios. Background Technology

[0002] With the deep advancement of the low-carbon strategy, the installed capacity of new energy power plants, including wind power, photovoltaic power, energy storage, and integrated wind-solar-storage systems, has continued to grow rapidly and has become a core component of the new power system. New energy power plants generally have characteristics such as complex equipment types, complex electrical topologies, significant influence of meteorological and geographical environments on operating conditions, and strong suddenness and chain reaction of fault evolution. Their safe and stable operation is highly dependent on the equipment knowledge, standardized operation skills, emergency response capabilities for sudden faults, and risk prediction capabilities of operation and maintenance personnel. The skill level of operation and maintenance personnel directly determines the power generation efficiency, equipment lifespan, and safety production baseline of the power plant.

[0003] With the rapid development of the new energy industry, the scale of new energy power plants such as wind power and photovoltaic power generation is constantly expanding, and the types of equipment are becoming increasingly complex, significantly increasing the difficulty of operation and maintenance. New energy power plants are typically located in remote areas with harsh environments, resulting in frequent equipment failures. These failures are diverse in type and complex in their evolution, placing extremely high demands on the emergency response capabilities of operation and maintenance personnel. Traditional operation and maintenance training mainly relies on classroom lectures, document learning, or simple two-dimensional simulation systems, which have the following prominent problems: Traditional training cannot accurately reproduce the complex geographical environment of new energy power plants, the spatial layout of equipment, and the impact of meteorological conditions on failure evolution. Operation and maintenance personnel cannot accumulate operational experience in near-realistic scenarios, leading to a significant gap between training effectiveness and actual needs. Existing simulation training systems mostly use static fault point settings, failing to simulate the complete evolution process of a fault from latency, outbreak, propagation, to recovery. Operation and maintenance personnel lack practical training in dealing with dynamic changes in faults, often resulting in slow reactions and improper handling when facing real faults.

[0004] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0005] In view of this, the present invention provides a three-dimensional GIS simulation training system for new energy power stations based on real-world scenarios to solve the aforementioned problems.

[0006] To solve the above problems, the specific technical solution adopted by the present invention is as follows:

[0007] A 3D GIS simulation training system for new energy power stations, built based on real-world scenarios, includes:

[0008] The new energy power station 3D scene unit is used to display the operating status information, fault location and display, and alarm information of each piece of equipment in the new energy power station based on the 3D scene data of the new energy power station in a pre-built 3D real scene model, and to collect the fault display results to form a fault case library.

[0009] The full-dimensional quantitative evaluation unit is used to conduct dynamic evolutionary fault simulation tests on the operation and maintenance personnel of new energy power stations based on a three-dimensional real scene model and fault case library. It also combines dynamic evolutionary game theory and interval type II trapezoidal fuzzy set to conduct full-dimensional quantitative evaluation of the simulation test results.

[0010] The Skills Profiling and Training Strategy Building Unit is used to build skills profiles and dynamic collaborative training strategies for operations and maintenance personnel based on the quantitative evaluation results of simulation tests.

[0011] Preferably, the full-dimensional quantitative evaluation unit includes:

[0012] The fault script construction module is used to construct a dynamic evolutionary fault script containing several fault rescue stages based on the fault case library by extracting historical fault data and meteorological-geographical correlation features of new energy power stations.

[0013] The behavioral dataset inversion module is used to collect operational behavior data of new energy power station operation and maintenance personnel in real time during the dynamic evolutionary fault simulation test through dynamic evolutionary fault scripts using virtual sensors pre-configured in the 3D real scene model; and to obtain a multi-dimensional operational behavior dataset containing explicit operational actions and implicit cognitive load by performing cognitive state inversion on the operational behavior data.

[0014] The mapping evaluation module is used to map multi-dimensional operational behavior datasets based on interval type II trapezoidal fuzzy sets, and to quantitatively evaluate the mapping results based on the optimization objectives of each fault rescue stage, forming a full-dimensional quantitative evaluation of the simulation test results.

[0015] Preferably, the method utilizes virtual sensors pre-configured in a 3D reality model to collect real-time operational behavior data of new energy power station operation and maintenance personnel during dynamic evolutionary fault simulation testing using dynamic evolutionary fault scripts; and obtains a multi-dimensional operational behavior dataset containing explicit operational actions and implicit cognitive load by performing cognitive state inversion on the operational behavior data, including:

[0016] By pre-deploying virtual sensors in a 3D real-world model, real-time data on the operation trajectory and operation command intervals of new energy power station operation and maintenance personnel during the dynamic evolutionary fault simulation test are collected, and a raw operation behavior dataset is constructed.

[0017] Based on the safety specifications and fault handling standards for the operation and maintenance of new energy power stations, the original operational behaviors of new energy power station operation and maintenance personnel are cognitively deconstructed to obtain competing cognitive patterns, and the benefit function of each cognitive pattern at different fault evolution nodes is defined.

[0018] Based on evolutionary game theory, the replication dynamic evolution equations of each cognitive mode are constructed. Combined with the time series data of the original operational behavior and the payoff function, the change trajectory of each cognitive mode is inverted through numerical solution, so as to obtain the decision mode transfer path of the operation and maintenance personnel of the new energy power station in the whole process of simulation test.

[0019] Based on the decision-making mode transition path, the dominant cognitive mode type of each fault evolution stage is extracted, and combined with the original operational behavior data, a multi-dimensional operational behavior dataset containing explicit operational actions and implicit cognitive load is generated.

[0020] Preferably, based on the safety specifications and fault handling standards for the operation and maintenance of new energy power stations, the original operational behaviors of new energy power station operation and maintenance personnel are cognitively deconstructed to obtain competing cognitive models, and the benefit functions of each cognitive model at different fault evolution nodes are defined, including:

[0021] Based on the fault case library, we determine the fault handling standards for each piece of equipment in the new energy power station. In conjunction with the operation and maintenance safety specifications of the new energy power station, we conduct qualitative analysis on the original operation behavior data and extract the dominant cognitive pattern types that compete with each other in the fault evolution process.

[0022] Based on the dominant cognitive pattern type, the strategy space and behavioral characteristics of each cognitive pattern in the fault handling process are defined respectively;

[0023] By analyzing the fault characteristics, environmental constraints, and operational consequences at different nodes during the fault evolution process, a benefit function for each cognitive mode at the corresponding node is constructed.

[0024] Preferably, the revenue function is based on the success rate of operation, response time, resource cost and security impact factor, and the revenue parameters of each node are calibrated based on historical failure case data and domain expert knowledge.

[0025] Preferably, the process of constructing dynamic evolution equations for each cognitive mode based on evolutionary game theory, and combining the original operational behavior time-series data and payoff functions, numerically solving and inverting the change trajectory of each cognitive mode, yields the decision mode transfer path of new energy power station operation and maintenance personnel throughout the simulation test process, including:

[0026] Based on the payoff function of each cognitive mode at the corresponding node, an evolutionary game payoff matrix with cognitive modes as the strategy set is constructed, and the game payoff matrix corresponding to each failure evolution node is determined.

[0027] Based on evolutionary game theory and combined with the payoff matrix of each fault evolution node, a dynamic evolution equation for the replication of cognitive patterns is established to describe the dynamic law of the change of the proportion of different cognitive patterns in the group of new energy power plant operation and maintenance personnel over time.

[0028] Feature extraction was performed on the time-series data of the original operational behavior, and the dominant cognitive patterns and the proportion distribution of cognitive patterns of the operation and maintenance personnel of the new energy power station at each time point were obtained by inversion using the Bayesian inference method.

[0029] Based on the payoff parameters in the game payoff matrix and combined with the distribution of cognitive patterns, the coefficients of the replication dynamic evolution equation are calibrated.

[0030] The calibration-processed replication dynamic evolution equations are numerically solved to form the change trajectory of each cognitive mode, so as to obtain the decision mode transfer path of the new energy power station operation and maintenance personnel in the entire simulation test process.

[0031] Preferably, the step of extracting features from the original operational behavior time-series data and inverting the dominant cognitive patterns and their proportional distribution at each time point using Bayesian inference methods includes the following steps:

[0032] Spatial and temporal features were extracted from the original operational behavior time series data, and a multidimensional behavioral feature vector reflecting cognitive load and decision-making patterns was constructed based on the extraction results.

[0033] Based on the standards for handling faults at new energy power stations, a probabilistic observation model between cognitive patterns and multidimensional behavioral feature vectors is established, and the conditional probability distribution function of behavioral feature vectors under each cognitive pattern is determined.

[0034] Using the payoff parameters in the game payoff matrix as prior information of the prior distribution, and combining them with the likelihood function, the Bayesian filtering algorithm is used to recursively estimate the time series data of the original operation behavior, so as to obtain the posterior probability distribution of the cognitive pattern of the operation and maintenance personnel at each time point.

[0035] The dominant cognitive mode at that moment is determined based on the maximum a posteriori criterion, and the real-time proportional distribution of each cognitive mode is extracted.

[0036] Preferably, the step of calibration of the coefficients of the replication dynamic evolution equation based on the payoff parameters in the game payoff matrix and in conjunction with the cognitive pattern proportion distribution includes:

[0037] Based on the payoff parameters in the game payoff matrix and the mathematical form of the replication dynamic evolution equation, the set of coefficients to be calibrated is determined.

[0038] The real-time proportional distribution of each cognitive mode is used as the observation data, and the objective function for coefficient calibration is constructed with the goal of minimizing the error between the proportional evolution trajectory of the cognitive mode simulated by the replicated dynamic evolution equation and the observation data.

[0039] Solving the objective function yields the coefficient calibration values ​​that best fit the simulated trajectory to the observed data.

[0040] Preferably, the mapping processing of the multi-dimensional operational behavior dataset based on interval type II trapezoidal fuzzy sets, and the quantitative evaluation of the mapping results based on the optimization objectives of each fault rescue stage, to form a full-dimensional quantitative evaluation of the simulation test results, includes:

[0041] Based on the various indicators in the multi-dimensional operational behavior dataset, we construct the interval type II trapezoidal fuzzy set membership function corresponding to each indicator, and map the measured value of each indicator to the fuzzy set to obtain the fuzzy membership vector of each indicator at different fault evolution stages.

[0042] Based on the optimization objectives of each stage of fault rescue, evaluation criteria corresponding to operation success rate, response time, resource cost and safety impact factors are extracted; and based on the preset evaluation criterion weights, a fuzzy comprehensive evaluation matrix is ​​constructed.

[0043] By combining the fuzzy membership vectors of each indicator with the evaluation criteria, the interval type II trapezoidal fuzzy comprehensive evaluation value of the operational behavior at each fault stage is obtained.

[0044] The interval type II trapezoidal fuzzy comprehensive evaluation value is quantified to obtain the operation performance score of the maintenance personnel at each fault stage. Combined with the preset time sequence weights of each fault stage, the comprehensive evaluation result of the entire simulation test process is calculated.

[0045] Preferably, the step of constructing interval-type trapezoidal fuzzy set membership functions for each indicator based on the various indicators in the multi-dimensional operational behavior dataset, and mapping the measured value of each indicator to the fuzzy set to obtain the fuzzy membership vector of each indicator at different fault evolution stages includes:

[0046] Based on the types of indicators in the multi-dimensional operational behavior dataset and the historical data distribution of each indicator at different fault evolution stages, and combined with the knowledge of new energy power station operation and maintenance experts, the numerical domain range of each indicator at each fault evolution stage is determined.

[0047] The numerical domain of each indicator at each fault evolution stage is divided into several fuzzy linguistic variables, and the principal membership function parameter of the corresponding interval type II trapezoidal fuzzy set is set for each fuzzy linguistic variable.

[0048] Based on the principal membership function parameters, construct the interval type II trapezoidal fuzzy set membership function for each index corresponding to different linguistic variables at each fault evolution stage;

[0049] The measured values ​​of each indicator in the multi-dimensional operational behavior dataset at the corresponding fault evolution stage are substituted into the interval type II trapezoidal fuzzy set membership function, and the fuzzy membership vector composed of the membership intervals of each linguistic variable is obtained by calculation.

[0050] The beneficial effects of this invention are as follows:

[0051] 1. This invention combines 3D real-scene modeling with dynamic evolutionary fault scripts to achieve high-fidelity reproduction of new energy power station operation and maintenance training scenarios and realistic simulation of fault evolution; it realizes the transformation from unified training to personalized and precise improvement, significantly enhancing the pertinence and effectiveness of training, and providing scientific and systematic technical support for the skill optimization and emergency response capability building of new energy power station operation and maintenance teams.

[0052] 2. This invention utilizes evolutionary game theory to establish a replicative dynamic evolution equation. Combined with a Bayesian filtering algorithm, the payoff parameters in the game payoff matrix are used as prior information. The posterior probability distribution of cognitive patterns at each moment is recursively derived from the behavioral time-series data. This allows for the extraction of dominant cognitive patterns and real-time proportion distributions, achieving in-depth mining from external behavior to internal cognition. Through coefficient calibration, the evolution equation is accurately fitted to the measured data, forming a decision-making pattern transition path. A multi-dimensional operational behavior dataset containing explicit operational actions and implicit cognitive load is generated. This not only dynamically tracks the changes in the cognitive state of maintenance personnel throughout the entire fault evolution process and accurately identifies the switching nodes and dominant tendencies of decision-making patterns, but also provides three-dimensional data support for subsequent fuzzy quantitative evaluation, which includes both behavioral trajectories and cognitive loads.

[0053] 3. This invention maps operational behavior data into fuzzy membership vectors, effectively encompassing measurement noise and cognitive ambiguity; it optimizes the target extraction evaluation criteria based on the fault stage and constructs a comprehensive matrix, making the evaluation more aligned with actual needs; through synthesis and defuzzification, it obtains performance scores for each stage and the overall results of the entire process, significantly improving the reliability and discriminativeness of the evaluation results, achieving a smooth transition from qualitative to quantitative, and providing a decision-making basis that is both accurate and robust for the construction of skill profiles. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0055] Figure 1 This is a flowchart illustrating the principle of a three-dimensional GIS simulation training system for new energy power stations built based on a real-world scenario, according to an embodiment of the present invention.

[0056] Figure 2 This is a principle block diagram of a three-dimensional GIS simulation training system for new energy power stations built based on a real scene according to an embodiment of the present invention;

[0057] Figure 3 This is a principle block diagram of a three-dimensional GIS simulation training system for new energy power stations built based on a real scene according to an embodiment of the present invention;

[0058] Figure 4 This is a flowchart illustrating the construction of a multi-dimensional operational behavior dataset in a three-dimensional GIS simulation training system for new energy power stations based on a real-world scenario, according to an embodiment of the present invention.

[0059] Figure 5 This is a hardware structure block diagram of the host device according to an embodiment of the present invention.

[0060] In the picture:

[0061] 1. Three-dimensional scene unit for new energy power stations; 2. Full-dimensional quantitative assessment unit; 3. Skill profile and training strategy construction unit. Detailed Implementation

[0062] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0063] The methods and embodiments provided in this application can be executed on a host device or a similar computing device. Taking running on a host device as an example, such as... Figure 5 As shown, the host device may include one or more ( Figure 5 Only one is shown in the diagram. The processor (which may include, but is not limited to, a microprocessor (MCU) or programmable logic device (FPGA), etc.) and storage for storing data are also shown. The host device may further include transmission devices for communication functions and input / output devices. Those skilled in the art will understand that... Figure 5 The structure shown is for illustrative purposes only and does not limit the structure of the host device described above. For example, the host device may also include components that are larger than... Figure 5The more or fewer components shown, or having the same Figure 5 The different configurations shown.

[0064] The memory can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the exception handling method in this embodiment. The processor executes various functional applications and data processing by running the computer program stored in the memory, thus implementing the above-described method. The memory may include high-speed random access memory (RAM) and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the host device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks (LANs), mobile communication networks, and combinations thereof.

[0065] Transmission devices are used to receive or send data over a network. Specific examples of the network described above may include a wireless network provided by a communication provider for the host device. In one example, the transmission device includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device may be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0066] According to an embodiment of the present invention, a three-dimensional GIS simulation training system for new energy power stations based on real-world scenarios is provided.

[0067] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figures 1-4 As shown, according to a first embodiment of the present invention, a three-dimensional GIS simulation training system for new energy power stations based on real-world scenarios is provided, comprising:

[0068] The three-dimensional scene unit 1 of the new energy power station is used to display the operating status information, fault location and display and alarm information of each equipment in the new energy power station based on the three-dimensional scene data of the new energy power station in advance, and to collect the fault display results to form a fault case library.

[0069] It should be noted that the 3D scene data for new energy power stations needs to be collected through methods such as UAV oblique photography, laser scanning, or BIM modeling based on design drawings, including geospatial data, equipment appearance and structural data, and texture information of photovoltaic arrays, wind turbines, substations, transformer substations, and transmission lines. A high-precision 3D reality model is then built based on a 3D GIS engine, and the 3D reality model is registered with the real geographic coordinate system.

[0070] This system utilizes industrial communication protocols such as OPCUA and Modbus TCP / IP to access real-time equipment operation data from the new energy power plant monitoring system. This data includes, but is not limited to: wind turbine speed, power generation, temperature, and vibration frequency; photovoltaic module voltage, current, and irradiance; and substation switch status and load data. Visual mapping is then implemented; for example, normal equipment operation is displayed in green, abnormal warnings in yellow, and fault shutdowns in red. Alternatively, the system can display the operating status, parameters, and system diagrams of generator sets, main transformers, and their auxiliary systems across multiple power plants using graphical, graphical, bar chart, and trend chart formats. This provides production managers and operators with intuitive, user-friendly, and real-time updated data, helping them understand the overall system operation and the relationships between various monitoring points.

[0071] When the SCADA system detects a device malfunction or a preset fault triggered by a simulation script, the system automatically receives the fault signal. It can automatically locate the corresponding device model in the 3D scene based on the preset unique identifier ID of the faulty device. The location of the faulty device is highlighted in the 3D scene using methods such as flashing highlights, pointing arrows, and light bar markers to assist maintenance personnel in quick identification. Alternatively, it can use thermal imaging data, telemetry data, etc., to create a digital twin of the device's temperature cloud map, monitoring the device's status and operation in real time, performing data analysis and trend prediction, and intuitively displaying data change trends and abnormal alarms, making the data easier to understand and analyze, and enabling timely problem detection and handling. Online device health status diagnosis aims to ensure the normal operation of equipment. Based on the establishment of an equipment health information database, it extracts, analyzes, and reorganizes monitoring data and fault diagnosis prediction data sent by personnel or robots during inspections using artificial neural networks. Effective data is then pushed to the digital twin platform for real-time display, aiming to provide personalized decision-making. Inspection administrators only need to click on a device in the equipment list to access the device's health diagnosis.

[0072] In addition, the construction of the fault case library requires the automatic recording of scene snapshots, operation logs of the entire fault handling process, fault handling results, meteorological-geographic correlation data, etc., during each simulation test or real fault. This information is then structured according to a preset case template to form standardized cases consisting of a five-tuple containing fault type, fault phenomenon, handling process, handling result, and environmental characteristics. These cases are stored in a relational database or time-series database and support multi-dimensional retrieval by fault type, time range, equipment type, and other dimensions.

[0073] Specifically, virtualizing the power generation equipment, substations, and terrain of new energy power plants allows for a comprehensive display of the entire plant. This enables three-dimensional monitoring of the plant's operation, a three-dimensional display of the internal mechanisms and operational status of the power generation equipment and substations, and facilitates simulation training for production personnel and external visits. Specifically, a wind farm and a photovoltaic power station are selected for a full-view display, followed by detailed models of various mechanical components inside the substation, achieving a panoramic view. A three-dimensional virtual power plant is established using reverse modeling technology to create a virtual power plant consistent with the physical plant, including full-plant visualization navigation and 3D visualization extensions. This virtual plant is then integrated into an existing 3D smart power plant platform for digital management of physical equipment assets in a 3D visualization format.

[0074] For example, for a 3D rendering of a wind farm, a 3D rendering engine can fuse satellite maps, elevation data, and oblique photogrammetry models of the site to realistically display the operating parameters, turbine locations, and turbine operating conditions of a single wind farm. Specifically, this includes:

[0075] (1) Real-time data acquisition: Real-time data of the wind farm is collected through sensors and monitoring equipment, including operating parameters such as wind speed, temperature, humidity, and nacelle temperature, and the status and performance of the wind turbine are monitored in real time. (2) Based on the geographical data and design drawings of the actual wind farm, a high-precision three-dimensional model is established, including the wind turbine tower, blades, nacelle and other parts, and corresponding to the actual scene. (3) Real-time data update: The real-time collected data is associated with the three-dimensional model of the wind turbine, and the status, data display and animation effects of the wind turbine in the scene are updated in real time to achieve real operating effect and data feedback. (4) Data visualization display: Real-time data is displayed through the three-dimensional scene. For example, the rotation speed and power output of the wind turbine blades are displayed in the form of animation, and the wind speed, wind direction and temperature are displayed in the form of charts or graphs. (5) Abnormal status monitoring and early warning: The operating status of the wind turbine is monitored through real-time data analysis and comparison, and real-time early warning is given for abnormal situations, such as excessive temperature and abnormal working status, so that corresponding measures can be taken in advance to avoid failure.

[0076] In addition, video monitoring of wind farms can be connected; by connecting to the streaming media protocols of various monitoring stations, the monitoring screen can be opened in the scene. Real-time video monitoring of wind farms can be realized. Operation and maintenance personnel can view the video screens of each monitoring point through the monitoring platform and promptly discover and handle abnormal situations. Specifically, it can achieve: (1) Wind farm optimization and simulation; based on actual data and the geographical environment of the wind farm, wind farm optimization and simulation can be carried out, including wind turbine position adjustment, blade angle optimization, etc., to improve power generation efficiency. (2) Historical data analysis: historical data can be replayed and analyzed, and the operating status of wind farms in different time periods can be compared and analyzed to help optimize operation and maintenance decisions and management. (3) User interaction and operation: a user interaction interface is provided, including visual operation tools, scene browsing and navigation functions, to facilitate users to view and analyze real-time and historical data of wind farms.

[0077] The full-dimensional quantitative evaluation unit 2 is used to conduct dynamic evolutionary fault simulation tests on the operation and maintenance personnel of new energy power stations based on a three-dimensional real scene model and fault case library, and to conduct full-dimensional quantitative evaluation of the simulation test results by combining dynamic evolutionary game theory and interval type II trapezoidal fuzzy set.

[0078] In a preferred embodiment, the full-dimensional quantitative evaluation unit 2 includes:

[0079] The fault script construction module is used to construct a dynamic evolutionary fault script containing several fault rescue stages based on the fault case library by extracting historical fault data and meteorological-geographical correlation features of new energy power stations.

[0080] It should be noted that, based on historical fault records stored in the fault case library, a dynamic fault script with temporal evolution and environmental coupling characteristics is generated by extracting the technical parameters of the fault itself and the meteorological-geographical correlation features at the time of the fault occurrence. This script contains multiple interrelated fault rescue stages, simulating the entire process of a real fault from occurrence, development, spread to final handling. Specifically: standardized fault case data is read from the fault case library constructed from the three-dimensional scene units of the new energy power station, and each case contains five-tuple information: fault type, fault phenomenon, handling process, handling result, and environmental characteristics; then, meteorological elements at the time of the fault occurrence are extracted from the environmental characteristic fields of the fault case library, including but not limited to: wind speed, wind direction, ambient temperature, humidity, irradiance, precipitation, and lightning activity index. Finally, the geographical location features of the faulty equipment are extracted from the geographic information system of the three-dimensional scene unit, including: terrain slope, altitude, distribution of surrounding obstacles, distribution of light and shadow, and soil resistivity. Then, meteorological and geographical features are spatiotemporally aligned to construct a joint meteorological-geographic feature vector. Based on historical fault case handling records, the entire fault process from occurrence to termination is divided into several evolutionary nodes with significant state differences. These evolutionary nodes are then merged and recombined to form several fault rescue stages. Each stage has clear starting conditions, ending conditions, stage objectives, and a set of typical operations. This generates a dynamic evolutionary fault script with temporal evolution characteristics and environmental coupling characteristics, used to simulate the entire process of a real fault from occurrence, development, propagation to final handling.

[0081] For example, in this embodiment, a preferred implementation example of how the fault script construction module constructs a dynamically evolving fault script is also provided, as follows:

[0082] Step A1: Based on the set of equipment objects, equipment connection relationships, equipment operating status information, fault stage records in the fault case library, and interlocking and isolation rules in the station operation and maintenance procedures in the 3D real scene model of the new energy power station, a unified constraint matrix is ​​formed by constructing equipment connection matrix, equipment state vector, mechanical interlock matrix, electrical interlock matrix, operation interlock matrix and permission boundary matrix, and then weighted and fused to output the integrated basic constraint model of equipment state-connection-constraint.

[0083] Specifically, step A1 includes:

[0084] A11: Extract all equipment objects within the site from the 3D reality model and construct a collection of equipment objects:

[0085] ;

[0086] in Indicates a collection of equipment at the station. Indicates the first A device object, This represents the total number of equipment objects, with a value ranging from 50 to 50,000 depending on the scale of the site.

[0087] It should be noted that this set serves as the base index for all subsequent matrix constructions.

[0088] A12: Construct a device connection matrix based on the physical connection relationships between devices: ;

[0089] in Represents the device connection matrix. Indicates equipment With equipment The physical connection between them. When and When there are primary loops, secondary loops, mechanical transmissions, control subordination, or energy transfer relationships. ;otherwise This matrix is ​​used to characterize the topological coupling relationships between devices.

[0090] A13: Construct a device state vector based on the current operating state of the device: ;

[0091] in Represents the device state vector. Indicates equipment The current status code. The status code is defined as a tiered value according to the type of equipment in the site. For example: 0 indicates separation or shutdown, 1 indicates connection or operation, 2 indicates maintenance lockout, 3 indicates grounding status, and 4 indicates permissioned isolation completed. This vector is used to characterize the operating status of the equipment in real time.

[0092] A14: Construct the mechanical interlocking matrix, electrical interlocking matrix, operational interlocking matrix, and permission boundary matrix respectively: ;

[0093] in, For mechanical interlocking matrix, Indicates device The action is affected by the equipment Mechanical position constraints; For electrical interlocking matrix, Indicates device The action is affected by the equipment Constrained by the energized state, protection state, or circuit state; To manipulate the locking matrix, Indicates device The actions are subject to the sequence of steps or the rules of the operation ticket; For the permission boundary matrix, Indicates device The operation must be performed on the equipment Execution can only proceed after verification of electricity, labeling, isolation, or permit confirmation has been completed.

[0094] A15: Weight and fuse the above four constraint matrices to construct a unified constraint matrix:

[0095] ;

[0096] in Represents the unified constraint matrix. To constrain the weighting coefficients, their values ​​range from 0.1 to 1.0, and they satisfy the following conditions: It should be noted that the weights are determined based on the type of power station, for example, in the case of a booster station. and A value of 0.3 to 0.4 is acceptable for internal wind turbine maintenance scenarios. A value of 0.25 to 0.35 is acceptable.

[0097] A16: Collect device objects Device connection matrix Device state vector With unified constraint matrix Combined, they form an integrated basic constraint model of equipment status, connection, and constraint:

[0098] ;

[0099] Step A2: Based on the basic constraint model And the set of fault rescue stages already divided in the fault case library. For each candidate operation within a fault stage, a pre-state vector, a post-state vector, and a stage target vector are constructed. The degree of matching between the device state and the pre-conditions is calculated using a state matching function. Whether the operation triggers interlocking or isolation conflicts is verified using a constraint satisfaction function. The overall result is a comprehensive executable decision value, outputting a set of executable operation units with mapping relationships between pre-constraints, post-states, and stage targets. .

[0100] Specifically, step A2 includes:

[0101] A21: Extract the set of divided fault rescue stages from the fault case database:

[0102] ;

[0103] in This represents the set of fault recovery phases. Indicates the first Each fault recovery phase The total number of stages is usually between 3 and 10.

[0104] It should be noted that each stage corresponds to a key node in the fault evolution process, such as the early warning incubation period, the fault outbreak period, the handling and control period, and the recovery and closing period. Each stage has clear starting conditions, ending conditions, and stage objectives.

[0105] A22: For each stage of troubleshooting Based on the fault characteristics, handling objectives, and typical operation records from historical cases at this stage, a set of candidate operations for this stage is extracted:

[0106] ;

[0107] in Representation phase The set of candidate operations within, Representation phase The Middle Candidate operations. This represents the number of candidate operations in this stage, typically ranging from 2 to 30.

[0108] It should be noted that the candidate operations cover all kinds of operations that may be performed in this stage, such as circuit breaker opening and closing, disconnecting switch operation, protection pressure plate activation and deactivation, parameter adjustment, emergency stop reset, etc.

[0109] A23: For each candidate operation Based on the set of device objects established in step 1 Construct its preceding state vector according to the index order: ;

[0110] in This represents a vector of state requirements that each device must meet before performing this operation, with a vector length of... Consistent with the total number of devices. The rules for determining the value are as follows:

[0111] When the value is -1, it indicates that the device This operation is irrelevant and does not constitute a prerequisite constraint; a value of 0 indicates that the device... It must be in a split or stopped state; a value of 1 indicates that the equipment is either in operation or stopped. It must be in the closed or running state; a value of 2 indicates that the device... It must be in maintenance lockout mode; a value of 3 or 4 indicates that the equipment... Grounding or permitted isolation must have been completed.

[0112] It should be noted that this vector specifies the device state boundary conditions that must be met before the operation is executed.

[0113] A24: For each candidate operation Construct its post-state vector:

[0114] ;

[0115] in This indicates the updated status of each device after the operation is performed. The same state coding system (0, 1, 2, 3, 4) as the previous state vector is used to indicate that the device has completed the operation. The target state to be achieved.

[0116] It should be noted that for devices unaffected by this operation, the corresponding position is set to -1 or retains the original state code, with the specific value determined based on the actual change in the device state caused by the operation.

[0117] A25: Construction Phase Target state vector: ;

[0118] in Representation phase The target state that each device should achieve at the end. The target state is encoded; for devices not involved, the value is -1.

[0119] It should be noted that this vector is used to determine whether a phase is complete. Phase advancement is only permitted when all relevant devices have reached the target state.

[0120] A26: Define the state matching function Used to measure the current device state vector Does the operation meet the requirements? Prerequisites: ;

[0121] in Indicates device The matching result of the operation preconditions is defined as follows:

[0122] ;

[0123] That is, the matching result is 1 when the device is irrelevant to the operation or its current state meets the prerequisites; otherwise, it is 0.

[0124] It should be noted that the state matching function The value range is [0,1], and the value is 1 if and only if all relevant devices meet the preconditions.

[0125] A27: Define constraint satisfaction function Used for verification operations Does it trigger constraint conflict between devices?

[0126] ;

[0127] in The unified constraint matrix in step 1 The Line number Column elements represent devices With equipment The strength of the constraint between them; As a conflict indicator variable, when the operation Will trigger device With equipment The value is 1 when there is a constraint conflict, and 0 otherwise. When any triggered strong constraint exists (i.e., ... When ), the maximum value of 1 is taken. This indicates that the operation is prohibited; when there are no constraint conflicts... This indicates that the operation meets the constraint requirements.

[0128] A28: Based on the combined state matching results and constraint satisfaction results, define the executable decision value for the operation:

[0129] ;

[0130] in Indicates operation An executable decision value at the current stage and under the current device state. If and only if When the operation satisfies all the requirements of the preceding equipment and does not trigger any interlocking or isolation conflicts, it is allowed to proceed; when When this occurs, it indicates that the operation cannot be executed directly and the operation order needs to be adjusted or other prerequisite operations need to be performed first.

[0131] A29: Combine each candidate operation with its associated preceding state vector, following state vector, stage target vector, and executable decision value to form an executable operation unit:

[0132] ;

[0133] By aggregating all executable operation units corresponding to all stages and all candidate operations, we obtain the set of executable operation units: ;

[0134] It should be noted that this set deepens the original set of typical operations organized only according to the fault stage into a set of executable operation units with clear pre-boundaries, post-results and stage goals, so that interlocking boundaries and isolation boundaries are formally written into the script expression structure, providing a basis for state updates and boundary propagation in subsequent steps.

[0135] Step A3: Based on the set of executable operation units With the initial device state vector The simulation process executes operation units hour by hour, updates device status, recalculates the executable decision value for each operation to generate an allowed operation indication vector, constructs a boundary blocking vector to represent isolation boundary propagation, calculates stage completion based on stage objectives, and outputs a state-boundary-allowed operation joint sequence that can be dynamically updated with each operation. .

[0136] Specifically, step A3 includes:

[0137] A31: Setting Simulation Propulsion Timing Variables The device state vector is used as the initial state: ;

[0138] in Indicates before the execution of the fault script ( The initial device state vector, This refers to the state vector defined in step 1.

[0139] It should be noted that the state at each subsequent moment is calculated based on the state at the previous moment and the change in state due to the operation performed.

[0140] A32: Assume at time... The operations and maintenance personnel select the executable operation unit to execute. (This operation unit has passed the executable determination in step 2, that is...) If this is executed, the device status will be updated as follows: ;

[0141] in Indicates time The device state vector, Indicates time The device state vector, For operation The resulting state increment vector.

[0142] A33: Calculate the state increment vector The One element: ;

[0143] In the formula Indicates device In performing operations The change in state afterwards The first state vector defined in step 2.4 One portion, For the current moment, the device The state is expressed intuitively by this formula: if the subsequent state is the same as the current state, the change is 0; otherwise, the change is the difference between the target state and the current state. (It should be noted that since the state code is a discrete value, the state can be directly set to the target code during the actual update, rather than arithmetic addition and subtraction. It is expressed in the form of difference here for the sake of uniform description.)

[0144] A34: After the device state is updated, based on the new state vector Recalculate the executable decision value for all candidate operation units (refer to steps 2.6 to 2.8) to obtain the allowed operation indication vector: ;

[0145] in Indicates time Permitted operation indicator vector, The total number of all candidate operation units. Indicates the first Each operation is allowed to be executed in the current state. This indicates that execution is prohibited.

[0146] A35: To express the propagation effect of the isolation boundary after the action is executed, a boundary blocking vector is constructed. First, at time... According to the mechanical interlocking matrix defined in step 1 Electrical interlocking matrix Operation of the locking matrix Permission Boundary Matrix Extraction devices Current status of whether it is subject to various types of blockades , , , The value can be either 0 or 1, where 1 indicates that the device is currently blocked and cannot be operated due to the corresponding type of interlocking or isolation rule.

[0147] Subsequently computing devices Overall lockdown value:

[0148] ;

[0149] In the formula This is the blockade strength coefficient, ranging from 0.2 to 1.0, and can be set according to the severity of different types of blockades (for example, mechanical blockades typically enforce physical prohibition and can take a higher value; permissive blockades may allow lifting under specific conditions and can take a lower value). When either type of blockade exists, Take the non-zero coefficient value corresponding to the block (in actual implementation, it can be directly determined as 1, and the coefficient can be used for subsequent display strength or priority sorting).

[0150] Define the boundary blocking vector: ;

[0151] when When, it indicates the equipment Having entered a no-crossing operational area, its blocked status should be displayed with a special marker in the 3D scene; when When this occurs, it indicates that the device is not blocked, and whether it can be operated can be determined based on the permission operation indication vector.

[0152] A36: Based on the stage target vector defined in step 2.5 Calculate the current stage At any moment Completion level: ;

[0153] in Indicates device The consistency between the current state and the stage target state is defined as:

[0154] ;

[0155] Contribute 1 when the device is irrelevant to the stage objective or its current state has already achieved the objective; otherwise, contribute 0. Stage Completion Rate The value range is [0,1], and the value is 1 if and only if all relevant devices have reached the target state.

[0156] A37: The script allows execution from the current stage when the following two conditions are met. Proceed to the next stage :

[0157] Current stage completion rate This means that all equipment has achieved the goals for this stage;

[0158] Next stage There exists at least one candidate operation. Executable decision value in the current state This means that there is an executable entry point for the next stage.

[0159] If the above conditions are not met, the script will remain in the current stage, and the operations and maintenance personnel can only continue to perform the operations allowed in this stage until the pre-isolation boundary and permission boundary meet the requirements and the stage is completed.

[0160] A38: Record the joint state-boundary-permitted operation information as a sequence at each moment during the entire simulation process: ;

[0161] in This represents the total number of simulation steps (i.e., the total number of operations executed during script execution). This sequence fully records the dynamic evolution of equipment status, the propagation of blocking boundaries, the real-time changes in the set of executable operations, and the completion status of each stage throughout the entire fault handling process, transforming the static fault stage description into a dynamic script that automatically tightens or releases execution boundaries based on the operation results.

[0162] Step A4: Based on the joint sequence In addition, the equipment models and interactive controls in the 3D GIS simulation interface of new energy power stations map the allowable operation indication vector at each moment to the enable state of the interface controls. Blocked equipment is visualized in the 3D scene with boundary display values. The engineering pass value is calculated in real time for the operation that the user attempts to perform to determine whether it is allowed to be written into the formal handling path. The formal handling path is bound to the stage sequence, the set of executable operation units, and the joint sequence to form a dynamic evolutionary fault simulation script with engineering executability constraints.

[0163] Specifically, step A4 includes:

[0164] A41: Combined sequence Every moment Allowed operation indicator vector Map the operation controls to the 3D GIS interactive interface and construct operation enable functions: ;

[0165] in Indicates the first An interactive control (such as a button, handle, or menu) at any time The enabling state; when When this happens, the control can be triggered to execute; when At this time, the control is locked and direct execution is not allowed.

[0166] It should be noted that this mapping enables dynamic control of the operation entry point by the script, ensuring that only operations permitted in the current state can be selected by the operations and maintenance personnel.

[0167] A42: For joint sequences Every moment Devices marked as blocked, based on the boundary blocking vector elements in Calculate its boundary display value in a 3D scene: ;

[0168] in Indicates device At any moment The boundary indicates intensity. This is the highlight factor for the device object's boundary in the interface, ranging from 0.5 to 1.5. Different values ​​can be preset based on the device's importance or the type of blocking. When When the device is in a restricted area, the system displays the device in a 3D scene using a preset color (such as red), a border, a light strip, or an isolation cover, visually conveying the restricted area to maintenance personnel.

[0169] A43: When maintenance personnel are constantly... Attempting to perform a certain operation When the corresponding control is clicked or triggered, the system calculates the pass value of the operation in real time: ;

[0170] in The enable state of the control defined in step 4.1, This represents the current stage completion percentage calculated in step 3.6. Project pass value. The value range of is [0,1]: if and only if and hour, This indicates that the operation satisfies both the current interlocking and isolation boundary requirements and the completed progress conditions of the current stage, and can be written into the formal handling path; when When this occurs, it indicates that although the operation may have business significance on the interface, it is not feasible under the current real-world project boundary. The system directly prohibits the operation and keeps the current stage unchanged.

[0171] A44: During the simulation test, record all successfully executed and satisfied results sequentially. The operation forms a formal handling procedure: ;

[0172] in Indicates the formal handling procedure. Indicates the first The operations that were allowed and successfully executed (in order of execution time). This indicates the total number of operations in the formal path.

[0173] It should be noted that this path fully records the sequence of fault handling steps actually completed by the operations and maintenance personnel under the constraints of the project boundaries.

[0174] A45: Formal handling procedures will be announced. With fault rescue phase sequence Executable operation unit set State-boundary-allowed operation joint sequence Binding is performed to create a dynamic evolutionary fault simulation script with engineering executability constraints: ;

[0175] in The final output is a dynamically evolving fault simulation script. Represents a stage sequence. Represents a set of executable operation units. Represents a joint sequence of state-boundary-permitted operations. This indicates the formal disposal path after the engineering boundary verification.

[0176] It should be noted that this script transforms the original static fault description, which was organized only by phenomena, into a dynamic simulation model that includes interlocking chains, isolation chains, permission chains, and actual handling paths. It can be directly used for loading and execution in a 3D GIS simulation training system.

[0177] The behavioral dataset inversion module is used to collect operational behavior data of new energy power station operation and maintenance personnel in real time during the dynamic evolutionary fault simulation test through dynamic evolutionary fault scripts using virtual sensors pre-configured in the 3D real scene model; and to obtain a multi-dimensional operational behavior dataset containing explicit operational actions and implicit cognitive load by performing cognitive state inversion on the operational behavior data.

[0178] In a preferred embodiment, the method utilizes virtual sensors pre-configured in a 3D real-world model to collect real-time operational behavior data of new energy power station operation and maintenance personnel during dynamic evolutionary fault simulation testing using dynamic evolutionary fault scripts; and obtains a multi-dimensional operational behavior dataset containing explicit operational actions and implicit cognitive load by performing cognitive state inversion on the operational behavior data.

[0179] By pre-deploying virtual sensors in a 3D real-world model, real-time data on the operation trajectory and operation command intervals of new energy power station operation and maintenance personnel during the dynamic evolutionary fault simulation test are collected, and a raw operation behavior dataset is constructed.

[0180] It should be noted that the virtual sensor includes a spatial position sensor embedded in the 3D scene, an operation command capture module, and a timestamp recording unit. When maintenance personnel interact with the 3D scene, it automatically records the 3D spatial coordinate sequence of their touch operation trajectory, the viewpoint switching path, the selected device object, and the type of control command issued (e.g., remote shutdown, parameter adjustment, reset operations). It also records the interval between the occurrence of a fault and the issuance of each operation command, as well as the time interval between adjacent commands. Finally, the collected raw data is stored and indexed in chronological order to form a raw operation behavior dataset containing spatiotemporal information and temporal characteristics.

[0181] Based on the safety specifications and fault handling standards for the operation and maintenance of new energy power stations, the original operational behaviors of new energy power station operation and maintenance personnel are cognitively deconstructed to obtain competing cognitive patterns, and the benefit function of each cognitive pattern at different fault evolution nodes is defined.

[0182] As a preferred implementation, the original operational behaviors of new energy power station maintenance personnel are cognitively deconstructed based on the safety specifications and fault handling standards for new energy power station operation and maintenance, resulting in competing cognitive patterns. The benefit functions for each cognitive pattern at different fault evolution nodes are defined as follows:

[0183] Based on the fault case library, we determine the fault handling standards for each piece of equipment in the new energy power station. In conjunction with the operation and maintenance safety specifications of the new energy power station, we conduct qualitative analysis on the original operation behavior data and extract the dominant cognitive pattern types that compete with each other in the fault evolution process.

[0184] It should be noted that the Safety Specifications for the Operation and Maintenance of New Energy Power Stations: encompassing industry standards, enterprise operating procedures and emergency plans, clarifies the standard handling procedures, safety prohibitions and priority principles for various faults, and serves as the benchmark for judging the compliance of operational behavior.

[0185] The qualitative analysis of the raw operational data employed a qualitative research method combining grounded theory and topic analysis. This method systematically coded and summarized the data, specifically including:

[0186] The raw operational behavior data is broken down item by item, and key behavioral characteristics are marked, such as strictly following the procedure sequence, skipping certain inspection steps based on past experience, and pressing the emergency stop button immediately when an alarm is triggered, thus forming initial concept labels.

[0187] Categorize concepts with similar characteristics to extract higher-level categories. For example, sequential operation and item-by-item checklist verification can be categorized as compliance tendency; operation based on memory of similar failures and workarounds can be categorized as experience matching tendency; and instinctive reactions and rapid interruptions can be categorized as stress response tendency.

[0188] A theoretical framework was constructed around the core categories, identifying three dominant cognitive modes that consistently exist and compete with each other during the failure evolution process. These dominant cognitive mode types include norm-compliant cognitive mode, experience-matching cognitive mode, and stress-response cognitive mode.

[0189] The compliance-oriented cognitive model involves making decisions by strictly adhering to operational procedures, fault handling standards, and safety regulations, with these procedures serving as the highest guiding principle. Its behavioral characteristics include: fixed and rigorous operating steps; the need to consult manuals or compare against standards during the decision-making process; moderate reaction speed; and an extremely low error rate within the scope of the procedures.

[0190] Experience-based cognitive models rely on an individual's or team's past experience in handling faults to make decisions, quickly selecting operational solutions by analogy with similar situations. Their behavioral characteristics include: flexible operational paths, fast decision-making speed, and an ability to extract effective patterns from experience; however, they may misjudge situations when faced with unfamiliar faults or sudden environmental changes.

[0191] The stress-response cognitive pattern involves making rapid responses based on instinct, intuition, or simple heuristic rules under time pressure, high workload, or sudden situations to reduce immediate risks. Its behavioral characteristics include extremely short reaction times, simple and direct operations, prioritizing the elimination of the fault source or activation of emergency protection, but potentially ignoring secondary aspects or subsequent impacts.

[0192] Based on the dominant cognitive pattern type, the strategy space and behavioral characteristics of each cognitive pattern in the fault handling process are defined respectively;

[0193] It should be noted that the strategy space refers to the set of all operational solutions that operations and maintenance personnel might take when facing a fault scenario under the guidance of a specific cognitive mode. This set is not an unlimited enumeration, but rather a summary and constraint based on actual handling records, operational procedures, and expert experience from a historical case library. For the three dominant cognitive modes, the strategy space is defined as follows:

[0194] The strategy space of the compliance-oriented cognitive model is centered on the standard operating procedures (SOPs) clearly defined in the safety specifications for the operation and maintenance of new energy power plants, equipment manufacturer technical manuals, and emergency plans. This space includes all operational sequences that comply with the procedures, such as: identifying fault types according to the alarm code lookup table; and strictly implementing safety isolation steps such as power outage, tagging, and voltage testing.

[0195] The strategy space of the experience-matching cognitive model is based on the successful handling experience accumulated by individual operations personnel or teams in similar past failures. This space includes practically proven workarounds, such as: skipping some redundant check steps based on experience to directly locate common failure points; and setting parameters by referring to the correction values ​​in historical cases rather than strictly following theoretical values.

[0196] The strategy space of the stress-response cognitive model is primarily focused on actions that can quickly interrupt the worsening of a fault or ensure the safety of personnel and equipment in an emergency. This space includes actions triggered by instinctive reactions or simple heuristic rules, such as: immediately pressing the emergency stop button or triggering a remote shutdown command; directly opening the fault circuit switch without first confirming the load transfer situation.

[0197] Furthermore, behavioral characteristics refer to the observable and quantifiable external behavioral patterns exhibited by operations and maintenance personnel during operations under specific cognitive patterns. These characteristics are extracted through operation trajectory data and command interval data collected by virtual sensors, and specifically include:

[0198] The behavioral characteristics of the norm-compliant cognitive model are: fixed operating steps, strict sequence, need to consult manuals or compare with standards in the decision-making process, moderate reaction speed, but extremely low error rate within the scope of the procedure.

[0199] The behavioral characteristics of experience-matching cognitive patterns are: the operation path is flexible, the decision-making speed is fast, and it is good at extracting effective patterns from experience, but it may make misjudgments when faced with unfamiliar faults or sudden environmental changes.

[0200] The behavioral characteristics of the stress-response cognitive pattern are: extremely short reaction time, simple and direct operation, priority to cut off the source of failure or activate emergency protection, but may ignore secondary links or subsequent effects.

[0201] By analyzing the fault characteristics, environmental constraints, and operational consequences at different nodes during the fault evolution process, a benefit function for each cognitive mode at the corresponding node is constructed.

[0202] It should be noted that the revenue function is based on the success rate of operation, response time, resource cost and security impact factor, and the revenue parameters of each node are calibrated based on historical failure case data and domain expert knowledge.

[0203] Specifically, the operational success rate refers to the probability of achieving the predetermined goals for this stage, such as fault isolation, parameter adjustment, and equipment restart, given a fault evolution node and cognitive mode. This probability is derived from the frequency statistics of successful handling of similar faults under the same cognitive mode in a historical case library, and can be dynamically adjusted by combining simulation test data. For example, during the outbreak of wind turbine pitch failure, the success rate of strictly following procedures under the compliance mode may be as high as 95%, while the success rate of direct reset under the stress response mode may only be 30%.

[0204] Response timeliness refers to the time elapsed from the triggering of a fault to the completion of a critical operation. It is measured by analyzing the time lag from the occurrence of the fault to the first valid instruction in the operation instruction interval data, as well as the completion time of subsequent critical operations. For example, during the fault propagation phase, experience-based matching patterns may be more than 20% faster than standard-compliant patterns, thus more effectively curbing the spread of the fault.

[0205] Resource costs refer to the human, material, and financial resources consumed during fault handling. These include: the cost of calling up spare parts, tool wear and tear, the cost of multi-person collaboration, and power loss due to system downtime. Resource costs can be normalized using handling cost data recorded in historical cases. For example, a stress-response model often involves simple emergency stops with relatively low resource consumption, but improper operation may trigger secondary faults, increasing subsequent costs.

[0206] The safety impact factor refers to the potential impact of operations on personal safety, equipment safety, and environmental safety. This factor is assigned a value based on the risk level classification in the safety regulations for the operation and maintenance of new energy power plants, such as levels 1 to 5, and further calibrated using expert experience to determine the risk coefficient for different operations. For example, the compliance-oriented model strictly follows safety procedures, resulting in the lowest safety impact factor; the stress-response model may overlook safety procedures, leading to increased safety risks.

[0207] Specifically, the payoff function is typically expressed as a weighted sum or product of the four dimensions mentioned above. The expression is:

[0208] ;

[0209] In the formula, Indicates the first Cognitive patterns in the first The profit value at each failure evolution node; To calculate the success rate of the operation, normalize to the [0,1] interval; Indicates the response time rating; Indicates resource cost score; Indicates the safety impact factor score; The weights for the corresponding dimensions are summed to one.

[0210] Table 1 lists the values ​​of the cognitive pattern benefit function parameters at each fault evolution node.

[0211] Table 1. Examples of parameters for the cognitive pattern benefit function at each fault evolution node.

[0212] Based on evolutionary game theory, the replication dynamic evolution equations of each cognitive mode are constructed. Combined with the time series data of the original operational behavior and the payoff function, the change trajectory of each cognitive mode is inverted through numerical solution, so as to obtain the decision mode transfer path of the operation and maintenance personnel of the new energy power station in the whole process of simulation test.

[0213] As a preferred implementation, the method of constructing dynamic evolution equations for each cognitive mode based on evolutionary game theory, and combining the original operational behavior time-series data and payoff functions, numerically solves and inverts the change trajectory of each cognitive mode to obtain the decision mode transfer path of the new energy power station operation and maintenance personnel throughout the simulation test process, including:

[0214] Based on the payoff function of each cognitive mode at the corresponding node, an evolutionary game payoff matrix with cognitive modes as the strategy set is constructed, and the game payoff matrix corresponding to each failure evolution node is determined.

[0215] It should be noted that, from an evolutionary game perspective, the operation and maintenance personnel of new energy power stations are considered as a group, and each individual in the group adopts a certain cognitive pattern as the dominant strategy at a specific moment. The payoff matrix describes the expected revenue that can be obtained by adopting a certain strategy when different individuals in the group adopt different strategies. Each element in this game payoff matrix is ​​a constructed payoff function.

[0216] Based on evolutionary game theory and combined with the payoff matrix of each fault evolution node, a dynamic evolution equation for the replication of cognitive patterns is established to describe the dynamic law of the change of the proportion of different cognitive patterns in the group of new energy power plant operation and maintenance personnel over time.

[0217] It should be noted that evolutionary game theory is used to establish a replicating dynamic evolution equation to describe the dynamic changes in the proportion of the three cognitive modes among the operations and maintenance personnel over time. The replicating dynamic equation serves as a bridge connecting static payoff comparisons with dynamic evolutionary trajectories. Specifically, it describes how the proportion of different strategies evolves over time within a group of boundedly rational individuals. Its core idea is that the growth rate of an individual employing a certain strategy in the group is equal to the difference between the current payoff of that strategy and the average payoff of the group. The proportion of strategies with payoffs higher than the average will increase, and vice versa. Specifically, this includes:

[0218] Assumption , , They represent The proportion of individuals in a group who consistently employ norm-compliant, experience-matching, and stress-response cognitive patterns satisfies the following:

[0219] ;

[0220] Regarding the first For each fault evolution node, the general form of the replicated dynamic equation is:

[0221] ;

[0222] In the formula, This indicates that when the group state is At that time, adopt strategies The individual's expected return; This represents the average expected return of the group, i.e. .

[0223] Feature extraction was performed on the time-series data of the original operational behavior, and the dominant cognitive patterns and the proportion distribution of cognitive patterns of the operation and maintenance personnel of the new energy power station at each time point were obtained by inversion using the Bayesian inference method.

[0224] As a preferred embodiment, the step of extracting features from the original operational behavior time-series data and inverting the dominant cognitive patterns and cognitive pattern proportion distribution of the new energy power station operation and maintenance personnel at each moment using a Bayesian inference method includes the following steps:

[0225] Spatial and temporal features were extracted from the original operational behavior time series data, and a multidimensional behavioral feature vector reflecting cognitive load and decision-making patterns was constructed based on the extraction results.

[0226] It should be noted that spatial feature extraction includes: operation path standardization index and operation sequence compliance. The operation path standardization index compares the actual operation trajectory with the ideal operation path specified in the fault handling standard and calculates the spatial similarity of the trajectory. The operation sequence compliance analyzes whether the execution order of operation instructions is consistent with the standard operating procedure and calculates the Kendall rank correlation coefficient between the actual order and the standard order.

[0227] The temporal feature extraction includes: reaction time lag feature, operation rhythm variation coefficient, and decision hesitation index. The reaction time lag feature is the time interval between the triggering of a fault node and the first valid operation command, characterizing the operator's alertness to the fault and the speed of decision initiation. The operation rhythm variation coefficient is the ratio of the standard deviation to the mean of the interval between adjacent operation commands, reflecting the stability of the operation rhythm. The decision hesitation index is calculated by statistically analyzing the number of times the viewpoint switches back and forth between the critical equipment and the control panel during the operation, as well as the time the mouse hovers without operation, characterizing the degree of hesitation in the decision-making process.

[0228] After normalizing the aforementioned spatial and temporal features, they are combined to form a multidimensional behavioral feature vector. , where m represents the number of feature vectors, and each dimension corresponds to a specific quantization metric.

[0229] Based on the standards for handling faults at new energy power stations, a probabilistic observation model between cognitive patterns and multidimensional behavioral feature vectors is established, and the conditional probability distribution function of behavioral feature vectors under each cognitive pattern is determined.

[0230] It should be noted that the probabilistic observation model uses a multidimensional probability distribution to characterize the distribution pattern of the behavioral feature vectors of operation and maintenance personnel under different cognitive modes. The probabilistic observation model consists of the mean vector, covariance matrix, and mixed weights corresponding to each cognitive mode. The parameter settings of the probabilistic observation model are obtained by maximum likelihood estimation or Bayesian estimation based on the labeled samples of traceable cognitive modes in the historical fault case library. For scenarios with sparse samples, the parameters are calibrated by combining the knowledge of operation and maintenance experts of new energy power stations through the analytic hierarchy process or the Delphi method. The training data of the probabilistic observation model mainly comes from real fault handling cases in the fault case library with clear cognitive modes through post-event interviews or expert annotations, as well as sample data with cognitive labels collected during simulation testing.

[0231] Specifically, the probabilistic observation model between cognitive patterns and multidimensional behavioral feature vectors is based on the following assumptions: the internal cognitive patterns (latent variables) of operations and maintenance personnel are manifested through their external operational behaviors (observable variables). However, under the same cognitive pattern, due to individual differences, environmental noise, and other factors, the observed behavioral characteristics exhibit random fluctuations. Therefore, it is necessary to establish a probabilistic mapping from latent variables to observed variables, i.e., a likelihood function. ,in, This represents the dominant cognitive pattern at time t. These represent norm-compliant cognitive patterns, experience-matching cognitive patterns, and stress-response cognitive patterns, respectively.

[0232] For the three cognitive modes, the conditional probability distribution of their behavioral feature vectors is determined respectively: the norm-compliant cognitive mode is a multidimensional normal distribution, with the mean vector biased towards the high norm and medium time lag region, and the covariance matrix reflects individual differences; the experience-matching cognitive mode has a mean vector biased towards the medium norm and short time lag region, and the covariance may be large; the stress-response cognitive mode has a mean vector biased towards the low norm and very short time lag region, and the covariance may show a skewed distribution.

[0233] Specific parameters are calibrated in the following way: from the historical failure case database, samples with clear markings or cognitive patterns that can be confirmed through post-event interviews are selected, and the mean and variance of behavioral characteristics under each pattern are statistically analyzed; for cases with sparse data, a prior distribution is set in combination with domain expert knowledge, and Bayesian updates are performed using subsequent observation data.

[0234] Using the payoff parameters in the game payoff matrix as prior information of the prior distribution, and combining them with the likelihood function, the Bayesian filtering algorithm is used to recursively estimate the time series data of the original operation behavior, so as to obtain the posterior probability distribution of the cognitive pattern of the operation and maintenance personnel at each time point.

[0235] Specifically, the Bayesian filtering algorithm is based on a hidden Markov model, whose state space consists of three cognitive modes: norm-compliant, experience-matching, and stress-response. The state transition matrix is ​​derived by discretizing the replication dynamic evolution equation after coefficient calibration. The observation model uses a multidimensional Gaussian distribution as the likelihood function, and its mean vector and covariance matrix are obtained by maximum likelihood estimation based on sample data with cognitive labels in the historical fault case library. For the sparse data dimension, the analytic hierarchy process is used to calibrate it by combining the knowledge of new energy power station operation and maintenance experts.

[0236] The Hidden Markov Model (HMM) is a dual stochastic process model whose structure consists of a hidden state sequence and an observation sequence. The hidden state sequence corresponds to the actual cognitive patterns of operators at different times, such as compliance, experience matching, and stress response. This sequence itself is not directly observable, but it follows the Markov property, that is, the hidden state at the current time depends only on the hidden state at the previous time. The observation sequence corresponds to the extracted multidimensional behavioral feature vectors, which are explicit behavioral data that can be directly measured. The core parameter settings of the Hidden Markov Model consist of three parts: first, the initial state probability distribution, which is obtained by transforming the payoff parameters in the game payoff matrix using the softmax function; second, the state transition probability matrix, which is obtained by numerically integrating and normalizing the calibrated replication dynamic evolution equation over a unit time interval; and third, the observation probability distribution (i.e., emission probability), which uses a multidimensional Gaussian distribution as the likelihood function. For each cognitive mode, its mean vector reflects the central trend of the typical behavioral characteristics under that mode, and the covariance matrix characterizes the correlation and fluctuation range between feature dimensions. Both of these parameters are obtained by maximum likelihood estimation based on sample data with cognitive labels in a historical failure case library. For sparse sample dimensions, expert knowledge is introduced through the analytic hierarchy process for calibration to ensure that the mapping relationship between each feature dimension and the cognitive mode is interpretable.

[0237] It should be noted that the payoff parameters in the game payoff matrix reflect the fitness of different cognitive modes at the current fault evolution node, and can serve as prior information on the distribution of cognitive modes. This is reasonable because at a specific fault evolution node, certain cognitive modes are more likely to be adopted by operations and maintenance personnel due to their higher payoffs; this tendency constitutes prior knowledge. By integrating the prior knowledge derived from game theory with real-time collected operational behavior data, a Bayesian filtering algorithm is used to dynamically infer the probability distribution of operations and maintenance personnel's cognitive modes at each time step. Specifically, the payoff parameters in the game payoff matrix reflect the relative fitness of different cognitive modes at a specific fault evolution node; a higher payoff mode means a greater likelihood of being adopted by operations and maintenance personnel at that node, and this theoretical tendency constitutes the mathematical basis of the prior information.

[0238] The Bayesian filtering algorithm utilizes prior information and an established likelihood function—the probability of observing the current behavioral feature vector under a given cognitive pattern—to recursively estimate the original operational behavior time-series data. During this recursion, based on the posterior probability distribution of the previous time step and the state transition patterns between cognitive patterns, the prior distribution for the current time step is predicted. When the operational behavior observation data for the current time step is obtained, the likelihood function is used to calculate the degree of matching between the observed data and each cognitive pattern, and the predicted prior distribution is corrected to obtain the posterior probability distribution for the current time step. This posterior distribution integrates the prior information provided by game theory and the evidence provided by real-time behavioral data, quantitatively characterizing the probability that the operations and maintenance personnel belong to the norm-compliant, experience-matching, and stress-response cognitive patterns at the current time step. Through time-by-time recursive estimation, the Bayesian filtering algorithm can dynamically track the evolution of the operations and maintenance personnel's cognitive state.

[0239] The dominant cognitive mode at that moment is determined based on the maximum a posteriori criterion, and the real-time proportional distribution of each cognitive mode is extracted.

[0240] It should be noted that after obtaining the posterior probability distribution of the cognitive modes of the operations and maintenance personnel at each time point, this step uses the maximum a posteriori criterion to extract the discretized dominant cognitive mode. The principle of this criterion is intuitive and consistent with decision-making logic: among the posterior probabilities of the three cognitive modes, the one with the largest value is selected as the dominant mode at the current time. For example, if the calculated posterior probability of norm compliance is 0.7, experience matching is 0.2, and stress response is 0.1 at a certain time, then the dominant cognitive mode of the operations and maintenance personnel at that time is determined to be norm compliance. At the same time, the real-time proportion distribution of each cognitive mode is also extracted, that is, the three probability values ​​in the posterior probability distribution are directly used as the proportion of the three cognitive modes at that time.

[0241] Based on the payoff parameters in the game payoff matrix and combined with the distribution of cognitive patterns, the coefficients of the replication dynamic evolution equation are calibrated.

[0242] In a preferred embodiment, the step of calibrating the coefficients of the replication dynamic evolution equation based on the payoff parameters in the game payoff matrix and in conjunction with the cognitive pattern proportion distribution includes:

[0243] Based on the payoff parameters in the game payoff matrix and the mathematical form of the replication dynamic evolution equation, the set of coefficients to be calibrated is determined.

[0244] Specifically, based on the specific mathematical form of the established replication dynamic evolution equation, all undetermined coefficients in the equation, excluding known benefit parameters, are identified. For example, if the equation adopts a standard form, only a global learning rate parameter may need to be calibrated; if the equation is extended to include an inertia term, inertia coefficients need to be calibrated; if the interaction strength between different modes is considered, multiple coupling coefficients need to be calibrated. The following principles should be followed when determining the coefficient set: the number of coefficients should not be too large to avoid overfitting; typically, the most critical factors are selected based on historical experience and theoretical analysis. Each coefficient should have a clear practical interpretation, such as the learning rate corresponding to the speed of experience accumulation by operations personnel, and the inertia coefficient corresponding to the strength of psychological set.

[0245] The real-time proportional distribution of each cognitive mode is used as the observation data, and the objective function for coefficient calibration is constructed with the goal of minimizing the error between the proportional evolution trajectory of the cognitive mode simulated by the replicated dynamic evolution equation and the observation data.

[0246] Specifically, given a set of candidate coefficients Substituting the equations into the established dynamic evolution equations for replication, and using the initial time proportions of the observed data as initial conditions, numerical integration is performed on the equations to obtain the simulated trajectory of cognitive pattern proportion evolution. This trajectory is a set of coefficients. The objective function is used to quantify the error between the simulated trajectory and the observed data.

[0247] Solving the objective function yields the coefficient calibration values ​​that best fit the simulated trajectory to the observed data.

[0248] Specifically, given the nonlinear and nonconvex characteristics of the objective function, and the fact that the coefficients to be calibrated are usually continuous variables, a genetic algorithm can be used to solve the problem. This involves: randomly generating an initial population within the range of values ​​for the coefficients to be calibrated; calculating the corresponding objective function value for each individual (i.e., a set of coefficient values), with smaller values ​​indicating better fitting; updating the population according to the rules of the selected algorithm to gradually decrease the objective function value; stopping the iteration when the maximum number of iterations is reached, the change in the objective function value is less than a threshold, or a satisfactory solution is found; and using the optimized set of coefficients as the calibration result.

[0249] The calibration-processed replication dynamic evolution equations are numerically solved to form the change trajectory of each cognitive mode, so as to obtain the decision mode transfer path of the new energy power station operation and maintenance personnel in the entire simulation test process.

[0250] It should be noted that numerically solving the calibrated replication dynamic evolution equations applies the data-calibrated dynamics to the entire simulation test process to generate a continuous trajectory of changes in the cognitive patterns of operations and maintenance personnel. Specifically, after coefficient calibration, the replication dynamic evolution equations become a mathematical model that can realistically reflect the evolutionary laws of the cognitive patterns of operations and maintenance personnel. The numerical solution process uses the distribution of cognitive patterns at the start of the simulation test as the initial condition, and combines the payment matrix switching rules of each node in the fault evolution process. Numerical integration is used to solve the equations time-by-time, thus obtaining continuous curves showing the changes in the proportions of the three cognitive patterns—normative compliance, experience matching, and stress response—over the entire time period from the start to the end of the test. These curves visually demonstrate which cognitive pattern gradually becomes dominant, which pattern is suppressed, and the dynamic process of transition between patterns at different stages of fault evolution. Based on these trajectories, the decision-making pattern transition path can be further extracted, that is, the key moments and node sequences of the switching of the dominant cognitive pattern can be identified, such as the transition from normative compliance to experience matching, or the stress response from experience matching to stress response. This transfer path reveals the inherent decision-making patterns of operations and maintenance personnel when facing dynamically evolving faults.

[0251] Based on the decision-making mode transition path, the dominant cognitive mode type of each fault evolution stage is extracted, and combined with the original operational behavior data, a multi-dimensional operational behavior dataset containing explicit operational actions and implicit cognitive load is generated.

[0252] It is important to note that extracting the dominant cognitive mode type for each stage of fault evolution based on the decision-making mode transition path and generating a multi-dimensional operational behavior dataset by combining it with the original operational behavior data is a crucial step in connecting the cognitive inversion results with the final evaluation data. Specifically, aligning the decision-making mode transition path with the stage divisions in the fault evolution script determines whether the dominant cognitive mode of the operations and maintenance personnel at each fault evolution node, such as the incubation period, outbreak period, and propagation period, is compliance-oriented, experience-matching, or stress-response-oriented. This dominant mode type reflects the main decision-making tendency of the operations and maintenance personnel at that stage. This cognitive-level information is then integrated with the collected original operational behavior data: on the one hand, explicit operational actions in the original operational behavior are retained, such as operational paths, operational sequences, and operational instruction content; on the other hand, the dominant cognitive mode type, the proportion distribution of cognitive modes, and the switching point information in the decision-making mode transition path are used as a quantitative representation of implicit cognitive load. For example, if the dominant mode is stress-response-oriented at a certain stage, it means that the operations and maintenance personnel are under high psychological and time pressure at that stage, and this pressure itself constitutes part of the implicit cognitive load. By integrating explicit operational actions and implicit cognitive load into the same dataset, a multi-dimensional operational behavior dataset is ultimately formed.

[0253] The mapping evaluation module is used to map multi-dimensional operational behavior datasets based on interval type II trapezoidal fuzzy sets, and to quantitatively evaluate the mapping results based on the optimization objectives of each fault rescue stage, forming a full-dimensional quantitative evaluation of the simulation test results.

[0254] As a preferred implementation, the mapping processing of the multi-dimensional operational behavior dataset based on the interval type-II trapezoidal fuzzy set, and the quantitative evaluation of the mapping results based on the optimization objectives of each fault rescue stage, to form a full-dimensional quantitative evaluation of the simulation test results, includes:

[0255] Based on the various indicators in the multi-dimensional operational behavior dataset, we construct the interval type II trapezoidal fuzzy set membership function corresponding to each indicator, and map the measured value of each indicator to the fuzzy set to obtain the fuzzy membership vector of each indicator at different fault evolution stages.

[0256] In a preferred embodiment, the step of constructing an interval-type trapezoidal fuzzy set membership function for each indicator based on the various indicators in the multi-dimensional operational behavior dataset, and mapping the measured value of each indicator to the fuzzy set to obtain the fuzzy membership vector of each indicator at different fault evolution stages includes:

[0257] Based on the types of indicators in the multi-dimensional operational behavior dataset and the historical data distribution of each indicator at different fault evolution stages, and combined with the knowledge of new energy power station operation and maintenance experts, the numerical domain range of each indicator at each fault evolution stage is determined.

[0258] It should be noted that the various indicators in the multi-dimensional operational behavior dataset, such as the operational path standardization index, reaction time lag characteristics, and operational rhythm variation coefficient, have different physical meanings and dimensions, and their value distributions differ significantly at different fault evolution stages. Therefore, the statistical distribution characteristics of each indicator at different stages are extracted from the historical fault case database, including the minimum, maximum, mean, standard deviation, and typical value range. Simultaneously, the statistical results are verified and corrected by incorporating the domain knowledge of new energy power plant operation and maintenance experts, removing outliers and supplementing reasonable estimates under data sparsity conditions. Through this data-driven approach combined with expert knowledge, the numerical domain range of each indicator at each fault evolution stage is ultimately determined, that is, the interval between the minimum and maximum possible values ​​of the indicator under normal conditions.

[0259] The numerical domain of each indicator at each fault evolution stage is divided into several fuzzy linguistic variables, and the principal membership function parameter of the corresponding interval type II trapezoidal fuzzy set is set for each fuzzy linguistic variable.

[0260] It should be noted that, based on the required level of granularity in the assessment, the value range of each indicator at each stage of fault evolution is divided into several linguistic variables. For example, it can be divided into three levels: low, medium, and high, or more precisely, four levels: excellent, good, medium, and poor. Each linguistic variable corresponds to an interval-type trapezoidal fuzzy set, which is characterized by the upper and lower membership functions of the principal membership function. The four vertex parameters of the trapezoidal function—the left starting point, left peak point, right peak point, and right ending point—determine the shape and position of the membership function. The parameters of the principal membership function are set based on: the typical value ranges corresponding to each linguistic variable (e.g., "low" corresponds to a smaller value range, and "high" corresponds to a larger value range), the degree of overlap between adjacent linguistic variables, and the degree of uncertainty in the expert's judgment of boundary values.

[0261] Based on the principal membership function parameters, construct the interval type II trapezoidal fuzzy set membership function for each index corresponding to different linguistic variables at each fault evolution stage;

[0262] It should be noted that for each indicator and each linguistic variable, an upper membership function and a lower membership function are constructed based on its trapezoidal vertex parameters.

[0263] The upper membership function describes the maximum membership degree, i.e., the degree to which a linguistic variable is completely belonged to it. It typically uses a standard trapezoidal function form, taking a value of 1 within the apex of the trapezoid and linearly decreasing to 0 on the slopes. The lower membership function describes the minimum membership degree, usually obtained by shrinking the upper membership function inwards, forming a band-shaped region between the upper and lower membership functions. This region represents the range of membership degree fluctuations caused by data uncertainty or cognitive differences. During construction, it is necessary to ensure that the upper and lower membership functions satisfy the basic properties of interval type II fuzzy sets, i.e., the lower membership function is never greater than the upper membership function, and both are convex functions.

[0264] The measured values ​​of each indicator in the multi-dimensional operational behavior dataset at the corresponding fault evolution stage are substituted into the interval type II trapezoidal fuzzy set membership function, and the fuzzy membership vector composed of the membership intervals of each linguistic variable is obtained by calculation.

[0265] It should be noted that for each indicator's measured value at a certain moment, the measured value is successively substituted into the interval type-2 trapezoidal fuzzy set membership function of each linguistic variable corresponding to the indicator in the current fault evolution stage, and the lower and upper membership degrees of the measured value to each linguistic variable are calculated respectively. For example, for the indicator of operational path standardization index, the membership interval for "excellent" can be calculated as [0.7, 0.9], the membership interval for "good" as [0.2, 0.4], and the membership intervals for "medium" and "poor" may both be [0, 0]. Combining these membership intervals according to the order of the linguistic variables yields the fuzzy membership vector of the indicator at that moment.

[0266] Based on the optimization objectives of each stage of fault rescue, evaluation criteria corresponding to operation success rate, response time, resource cost and safety impact factors are extracted; and based on the preset evaluation criterion weights, a fuzzy comprehensive evaluation matrix is ​​constructed.

[0267] It should be noted that, based on the optimization objectives of each fault recovery stage, corresponding evaluation criteria are extracted from four core dimensions: operational success rate corresponds to operational effectiveness criterion, response timeliness corresponds to time efficiency criterion, resource cost corresponds to economic criterion, and safety impact factor corresponds to safety criterion. Each criterion needs to clearly define its meaning, measurement method, and correlation with the fault stage objectives. Then, the analytic hierarchy process (AHP) or entropy weight method is used to determine the weight of each evaluation criterion in the corresponding fault stage. The weight reflects the relative importance of each criterion at that stage. For example, during the fault outbreak phase, the weights of response timeliness and safety may be higher; while during the recovery phase, the weight of resource cost may increase. Finally, based on the fuzzy membership vectors of each indicator and the correspondence between each indicator and the evaluation criteria, a fuzzy comprehensive evaluation matrix is ​​constructed. The rows of this matrix correspond to each evaluation criterion, the columns correspond to each indicator, and the matrix elements represent the importance or contribution coefficient of each indicator under a given criterion.

[0268] By combining the fuzzy membership vectors of each indicator with the evaluation criteria, the interval type II trapezoidal fuzzy comprehensive evaluation value of the operational behavior at each fault stage is obtained.

[0269] It should be noted that the synthesis operation adopts a fuzzy comprehensive evaluation method, typically using a weighted average or main factor prominence synthesis operator. For interval type II trapezoidal fuzzy sets, the synthesis operation needs to process membership intervals rather than precise values. Therefore, interval arithmetic rules are used: the fuzzy membership intervals of each indicator are multiplied and added between intervals and the corresponding weight coefficients in the evaluation matrix, aggregating layer by layer to obtain the interval type II fuzzy comprehensive value under each evaluation criterion, and then further aggregating to obtain the comprehensive evaluation value for the entire fault stage. The synthesis result is still an interval type II trapezoidal fuzzy set, with its upper and lower membership functions obtained from interval operations, preserving the uncertainty information in the entire evaluation process.

[0270] The interval type II trapezoidal fuzzy comprehensive evaluation value is quantified to obtain the operation performance score of the maintenance personnel at each fault stage. Combined with the preset time sequence weights of each fault stage, the comprehensive evaluation result of the entire simulation test process is calculated.

[0271] It should be noted that the interval type II fuzzy set is converted into a precise numerical value using order reduction and defuzzification methods. Specifically, the interval type II fuzzy set can be reduced to a type I fuzzy set; defuzzification uses the maximum mean method to extract the precise score representing the comprehensive evaluation value from the reduced fuzzy set. This score reflects the operational performance of the maintenance personnel during this fault phase, with a higher score indicating better performance.

[0272] In this process, based on the importance or duration of each fault evolution node throughout the simulation test, a time-series weight is preset for each stage; for example, the burst option has a higher weight, and the latent option has a lower weight. The operational performance score of each stage is weighted and summed with the corresponding time-series weight to obtain the comprehensive evaluation result of the entire simulation test process. Furthermore, since interval type II fuzzy sets preserve uncertainty information, the confidence interval or uncertainty measure of the evaluation result can be given by analyzing the width of the membership interval. This ensures that the final evaluation report includes not only a quantitative score but also an explanation of the reliability of the evaluation results.

[0273] Unit 3, Skill Profile and Training Strategy Construction, is used to construct skill profiles and dynamic collaborative training strategies for operations and maintenance personnel based on the quantitative evaluation results of simulation tests.

[0274] It should be noted that, based on the operational performance scores at each fault stage and the comprehensive evaluation results of the entire process, combined with a multi-dimensional operational behavior dataset, a skill profile of operations and maintenance personnel is constructed from multiple dimensions. This skill profile not only includes explicit indicators such as traditional operational proficiency and procedural mastery, but also incorporates implicit features obtained through cognitive inversion, such as dominant cognitive pattern preferences, the rationality of decision-making pattern transition paths, and cognitive load tolerance under time pressure. This skill profile is presented in the form of radar charts or feature vectors, which can intuitively show the strengths and weaknesses of operations and maintenance personnel, providing a precise basis for the formulation of subsequent training strategies. Building on this, this unit further constructs a dynamic collaborative training strategy: Dynamism is reflected in the fact that the training content is not static, but automatically retrieves targeted fault scripts from the fault case library based on the weaknesses identified in the skill profile, and adjusts the simulation difficulty and evolution path to achieve precise training that "fills in the gaps"; collaboration is reflected in the system's ability to simultaneously analyze the skill profiles of multiple operations and maintenance personnel, identify the weaknesses and complementary relationships in the overall team capability structure, and thus recommend the optimal team combination or conduct collaborative drill scripts to improve overall emergency response capabilities. It has completed a full closed loop from data collection, cognitive inversion, quantitative assessment to capability profiling and precise training, making simulation training a truly effective tool for improving the practical capabilities of new energy power station operation and maintenance personnel.

[0275] According to a second embodiment of the present invention, a training method for three-dimensional GIS simulation of new energy power stations based on real-world scenarios is provided, comprising the following steps:

[0276] S1. Based on the three-dimensional scene data of the new energy power station, a three-dimensional real scene model is pre-built to display the operating status information, fault location and display, and alarm information of each piece of equipment in the new energy power station, and the fault display results are collected to form a fault case library.

[0277] S2. Based on a 3D real-scene model and a fault case library, dynamic evolutionary fault simulation tests are conducted on the operation and maintenance personnel of new energy power stations. Combined with dynamic evolutionary game theory and interval type II trapezoidal fuzzy set, the simulation test results are quantitatively evaluated in all dimensions.

[0278] S3. Based on the quantitative evaluation results of simulation tests, construct skill profiles for operation and maintenance personnel and dynamic collaborative training strategies.

[0279] According to a third embodiment of the present invention, an electronic device is provided, the electronic device comprising: one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to perform the steps in any of the above method embodiments.

[0280] According to a fourth embodiment of the present invention, a computer-readable storage medium is provided, wherein a computer program is stored in the computer-readable storage medium, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to perform the steps in any of the above method embodiments.

[0281] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0282] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0283] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A 3D GIS simulation training system for new energy power stations based on real-world scenarios, characterized in that, include: The new energy power station 3D scene unit is used to display the operating status information, fault location and display, and alarm information of each piece of equipment in the new energy power station based on the 3D scene data of the new energy power station in a pre-built 3D real scene model, and to collect the fault display results to form a fault case library. The full-dimensional quantitative evaluation unit is used to conduct dynamic evolutionary fault simulation tests on the operation and maintenance personnel of new energy power stations based on a three-dimensional real scene model and fault case library. It also combines dynamic evolutionary game theory and interval type II trapezoidal fuzzy set to conduct full-dimensional quantitative evaluation of the simulation test results. The Skills Profiling and Training Strategy Building Unit is used to build skills profiles and dynamic collaborative training strategies for operations and maintenance personnel based on the quantitative evaluation results of simulation tests.

2. The new energy power station 3D GIS simulation training system based on a real-world scenario as described in claim 1, characterized in that, The comprehensive quantitative evaluation unit includes: The fault script construction module is used to construct a dynamic evolutionary fault script containing several fault rescue stages based on the fault case library by extracting historical fault data and meteorological-geographical correlation features of new energy power stations. The behavioral dataset inversion module is used to collect operational behavior data of new energy power station operation and maintenance personnel in real time during the dynamic evolutionary fault simulation test through dynamic evolutionary fault scripts using virtual sensors pre-configured in the 3D real scene model; and to obtain a multi-dimensional operational behavior dataset containing explicit operational actions and implicit cognitive load by performing cognitive state inversion on the operational behavior data. The mapping evaluation module is used to map multi-dimensional operational behavior datasets based on interval type II trapezoidal fuzzy sets, and to quantitatively evaluate the mapping results based on the optimization objectives of each fault rescue stage, forming a full-dimensional quantitative evaluation of the simulation test results.

3. The new energy power station 3D GIS simulation training system based on a real-world scenario, as described in claim 2, is characterized in that... The system utilizes virtual sensors pre-configured in a 3D real-world model to collect real-time operational data from new energy power station maintenance personnel during dynamic evolutionary fault simulation testing using dynamic evolutionary fault scripts. Furthermore, by performing cognitive state inversion on the operational behavior data, a multi-dimensional operational behavior dataset containing explicit operational actions and implicit cognitive load was obtained, including: By pre-deploying virtual sensors in a 3D real-world model, real-time data on the operation trajectory and operation command intervals of new energy power station operation and maintenance personnel during the dynamic evolutionary fault simulation test are collected, and a raw operation behavior dataset is constructed. Based on the safety specifications and fault handling standards for the operation and maintenance of new energy power stations, the original operational behaviors of new energy power station operation and maintenance personnel are cognitively deconstructed to obtain competing cognitive patterns, and the benefit function of each cognitive pattern at different fault evolution nodes is defined. Based on evolutionary game theory, the replication dynamic evolution equations of each cognitive mode are constructed. Combined with the time series data of the original operational behavior and the payoff function, the change trajectory of each cognitive mode is inverted through numerical solution, so as to obtain the decision mode transfer path of the operation and maintenance personnel of the new energy power station in the whole process of simulation test. Based on the decision-making mode transition path, the dominant cognitive mode type of each fault evolution stage is extracted, and combined with the original operational behavior data, a multi-dimensional operational behavior dataset containing explicit operational actions and implicit cognitive load is generated.

4. The new energy power station 3D GIS simulation training system based on a real-world scenario as described in claim 3, characterized in that, Based on the safety specifications and fault handling standards for the operation and maintenance of new energy power stations, the original operational behaviors of new energy power station operation and maintenance personnel are cognitively deconstructed to obtain competing cognitive patterns. The payoff functions for each cognitive pattern at different fault evolution nodes are defined, including: Based on the fault case library, we determine the fault handling standards for each piece of equipment in the new energy power station. In conjunction with the operation and maintenance safety specifications of the new energy power station, we conduct qualitative analysis on the original operation behavior data and extract the dominant cognitive pattern types that compete with each other in the fault evolution process. Based on the dominant cognitive pattern type, the strategy space and behavioral characteristics of each cognitive pattern in the fault handling process are defined respectively; By analyzing the fault characteristics, environmental constraints, and operational consequences at different nodes during the fault evolution process, a benefit function for each cognitive mode at the corresponding node is constructed.

5. The new energy power station 3D GIS simulation training system based on a real-world scenario, as described in claim 3, is characterized in that... The revenue function is based on the success rate of operation, response time, resource cost, and security impact factor, and the revenue parameters of each node are calibrated based on historical failure case data and domain expert knowledge.

6. The new energy power station 3D GIS simulation training system based on a real-world scenario as described in claim 3, characterized in that, The aforementioned dynamic evolution equations for replicating each cognitive mode are constructed based on evolutionary game theory. Combined with time-series data of original operational behaviors and payoff functions, the change trajectories of each cognitive mode are numerically solved and inverted. This yields the decision-making mode transition paths of new energy power plant operation and maintenance personnel throughout the simulation testing process, including: Based on the payoff function of each cognitive mode at the corresponding node, an evolutionary game payoff matrix with cognitive modes as the strategy set is constructed, and the game payoff matrix corresponding to each failure evolution node is determined. Based on evolutionary game theory and combined with the payoff matrix of each fault evolution node, a dynamic evolution equation for the replication of cognitive patterns is established to describe the dynamic law of the change of the proportion of different cognitive patterns in the group of new energy power plant operation and maintenance personnel over time. Feature extraction was performed on the time-series data of the original operational behavior, and the dominant cognitive patterns and the proportion distribution of cognitive patterns of the operation and maintenance personnel of the new energy power station at each time point were obtained by inversion using the Bayesian inference method. Based on the payoff parameters in the game payoff matrix and combined with the distribution of cognitive patterns, the coefficients of the replication dynamic evolution equation are calibrated. The calibration-processed replication dynamic evolution equations are numerically solved to form the change trajectory of each cognitive mode, so as to obtain the decision mode transfer path of the new energy power station operation and maintenance personnel in the entire simulation test process.

7. The new energy power station 3D GIS simulation training system based on a real-world scenario as described in claim 6, characterized in that, The process of extracting features from the original operational behavior time-series data and using Bayesian inference to invert and obtain the dominant cognitive patterns and their proportional distribution at each moment of the new energy power station operation and maintenance personnel includes the following steps: Spatial and temporal features were extracted from the original operational behavior time series data, and a multidimensional behavioral feature vector reflecting cognitive load and decision-making patterns was constructed based on the extraction results. Based on the standards for handling faults at new energy power stations, a probabilistic observation model between cognitive patterns and multidimensional behavioral feature vectors is established, and the conditional probability distribution function of behavioral feature vectors under each cognitive pattern is determined. Using the payoff parameters in the game payoff matrix as prior information of the prior distribution, and combining them with the likelihood function, the Bayesian filtering algorithm is used to recursively estimate the time series data of the original operation behavior, so as to obtain the posterior probability distribution of the cognitive pattern of the operation and maintenance personnel at each time point. The dominant cognitive mode at that moment is determined based on the maximum a posteriori criterion, and the real-time proportional distribution of each cognitive mode is extracted.

8. The new energy power station 3D GIS simulation training system based on a real-world scenario as described in claim 6, characterized in that, The process of calibrating the coefficients of the replication dynamic evolution equation based on the payoff parameters in the game payoff matrix and in conjunction with the cognitive pattern proportion distribution includes: Based on the payoff parameters in the game payoff matrix and the mathematical form of the replication dynamic evolution equation, the set of coefficients to be calibrated is determined. The real-time proportional distribution of each cognitive mode is used as the observation data, and the objective function for coefficient calibration is constructed with the goal of minimizing the error between the proportional evolution trajectory of the cognitive mode simulated by the replicated dynamic evolution equation and the observation data. Solving the objective function yields the coefficient calibration values ​​that best fit the simulated trajectory to the observed data.

9. The new energy power station 3D GIS simulation training system based on a real-world scenario as described in claim 8, characterized in that, The method of mapping the multi-dimensional operational behavior dataset based on interval type II trapezoidal fuzzy sets, and using the optimization objectives of each fault rescue stage as a benchmark, quantitatively evaluates the mapping results to form a comprehensive quantitative evaluation of the simulation test results, including: Based on the various indicators in the multi-dimensional operational behavior dataset, we construct the interval type II trapezoidal fuzzy set membership function corresponding to each indicator, and map the measured value of each indicator to the fuzzy set to obtain the fuzzy membership vector of each indicator at different fault evolution stages. Based on the optimization objectives of each stage of fault rescue, evaluation criteria corresponding to operation success rate, response time, resource cost and safety impact factors are extracted; and based on the preset evaluation criterion weights, a fuzzy comprehensive evaluation matrix is ​​constructed. By combining the fuzzy membership vectors of each indicator with the evaluation criteria, the interval type II trapezoidal fuzzy comprehensive evaluation value of the operational behavior at each fault stage is obtained. The interval type II trapezoidal fuzzy comprehensive evaluation value is quantified to obtain the operation performance score of the maintenance personnel at each fault stage. Combined with the preset time sequence weights of each fault stage, the comprehensive evaluation result of the entire simulation test process is calculated.

10. The new energy power station 3D GIS simulation training system based on a real-world scenario, as described in claim 9, is characterized in that... The step involves constructing interval-type trapezoidal fuzzy set membership functions for each indicator based on various indicators in the multi-dimensional operational behavior dataset, and mapping the measured values ​​of each indicator onto the fuzzy set to obtain the fuzzy membership vectors of each indicator at different fault evolution stages, including: Based on the types of indicators in the multi-dimensional operational behavior dataset and the historical data distribution of each indicator at different fault evolution stages, and combined with the knowledge of new energy power station operation and maintenance experts, the numerical domain range of each indicator at each fault evolution stage is determined. The numerical domain of each indicator at each fault evolution stage is divided into several fuzzy linguistic variables, and the principal membership function parameter of the corresponding interval type II trapezoidal fuzzy set is set for each fuzzy linguistic variable. Based on the principal membership function parameters, construct the interval type II trapezoidal fuzzy set membership function for each index corresponding to different linguistic variables at each fault evolution stage; The measured values ​​of each indicator in the multi-dimensional operational behavior dataset at the corresponding fault evolution stage are substituted into the interval type II trapezoidal fuzzy set membership function, and the fuzzy membership vector composed of the membership intervals of each linguistic variable is obtained by calculation.

Images